摘要
针对使用贪心法、遗传算法等方法实现选择性神经网络集成时出现的“局部最小点”和“过拟合”问题,提出了一类基于随机梯度法的选择性神经网络二次集成方法。理论分析和实验表明,与上述选择性神经网络集成方法相比,该方法易于实现且效果明显。
In the application of greedy method or genetic algorithm for the selection of the components of neural network ensembles, local minima and over fitting problems occur frequently. To solve such problems, a kind of method of two-level stochastic gradient-based selective neural network ensembles is proposed in this paper. Theoretical analyses and experimental results show that the method is easy to be constructed and performs well.
出处
《计算机工程》
CAS
CSCD
北大核心
2004年第16期133-135,159,共4页
Computer Engineering
关键词
神经网络集成
二次集成
贪心法
随机梯度法
Neural network ensembles
Two-level ensemble
Greedy method
Stochastic gradient method